The Order Kalshi Wrote, Washington's Amended Injunction and the Cross-Jurisdictional Impact of Consented Architecture of State Enforcement

National Prediction Market Litigation Architecture: How a Contempt Settlement Became a Portable State Enforcement Protocol

The Dispute in Plain Terms

Kalshi operates an online exchange where customers buy contracts that pay out based on real-world events — which team wins a game, who wins an election, whether a public figure says a particular word. A federal agency, the Commodity Futures Trading Commission, licenses the exchange and treats the products as financial derivatives.

Roughly a dozen states treat the same products as gambling. Their argument is simple: a contract paying out on a point spread is a sports bet regardless of what the seller calls it, and offering one without a state license breaks state law.

The disagreement has produced more than twenty state lawsuits and enforcement actions since early 2025, and courts have split. A federal appeals court in Philadelphia sided with Kalshi in April 2026. Federal and state courts in Ohio, Maryland, Connecticut, New York, Utah, and Wisconsin have ruled the other way.

The federal agency has gone further, suing states directly to stop them from enforcing their own gambling laws.

No court has settled the question nationally, and a controlling judicial answer is unlikely before 2027.

Full analysis, comparative state architecture, and the complete prediction registry: https://www.mindcast-ai.com/p/wa-kalshi-injunction

Related MindCast works: Kalshi Loses Federal Forum — The Washington Remand Order | How the CFTC's Missing "Gaming" Definition Is Losing the Preemption War | New York's $36 Billion Kalshi Case — The One Claim No CFTC Rule Can Erase

What Happened on August 12

A Washington state judge in Seattle entered three orders in one afternoon. The first bans Kalshi from offering seven categories of contracts to anyone in Washington — sports, elections, politics, culture, entertainment, technology and science, and contracts on what public figures will say. Four categories survive: commodities, climate, economics, and finance.

The second order refuses Kalshi permission to keep two exhibits confidential. The third denies Kalshi's request to pause the ban while it appeals.

The seven-category scope drew national coverage. Almost nobody examined the enforcement mechanism, which is the more consequential development.

The Central Finding

Washington did not design the compliance system now operating against Kalshi. Kalshi agreed to that system three weeks earlier, in a different state, to avoid a scheduled contempt hearing.

On July 23, Kalshi and Nevada gaming regulators filed a joint settlement. Kalshi accepted five terms: a named location-blocking vendor, a $120,000-per-day penalty for missing the deadline, an option to file a sworn explanation instead of paying, court-set penalties if that explanation shows insufficient effort, and a promise to share its technical progress reports with other states.

Washington's August 12 order reproduces all five terms and adds itself to the reporting list.

Two consequences follow. Kalshi retains every right to appeal, but arguing that the vendor, the schedule, or the penalty is unworkable means arguing against an agreement it signed weeks earlier. And the reporting requirement compounds — Nevada's version named Michigan, Washington's names Michigan and Nevada, and a fourth state can demand the same term at no cost while pointing to three states already receiving it.

A remedy a court invents invites appellate correction. A remedy assembled from a defendant's own settlement terms resists it.

Why the Federal Question Matters Less Than It Appears

Most coverage frames this litigation as one binary question: does federal law override state gambling law?

Washington's order does not depend on winning that question nationally. Washington won it in front of the judge hearing the case, then relied on three of its own statutes for the actual liability — the state gambling act, the consumer protection act, and a statute allowing recovery of gambling losses.

The federal agency has a pending rulemaking that would define "gaming" for the first time. A finished definition reaches none of Washington's three claims. The court held that Kalshi does not qualify for an exemption written into Washington's gambling statute, which is a question of Washington law no federal agency can answer.

A separate federal regulation now works against Kalshi as well. Federal rules bar a licensed exchange from listing contracts referencing activity "unlawful under any State law." Kalshi cites that provision as proof of exclusive federal control. The Washington judge read the same words as a federal instruction to consult state law first — and no federal definition of gambling can remove a clause that points to state law by its own terms.

One limit deserves emphasis. Federal preemption doctrine is a different instrument, and a controlling appellate ruling could still bar Washington from enforcing its statute. Protection from the rulemaking is not protection from preemption.

The Forecast

MindCast forecasts through simulation rather than commentary. The engine builds behavioral models of each material actor, plays opposing models against each other under pressure, tests every causal claim against an integrity standard, and releases only forecasts whose evidence chain holds.

MindCast ran the frozen record through two independent LLM-based foresight executions, each using eleven actor models. The systems agreed on the governing mechanism but produced distinct predictions and route probabilities. MindCast preserves those differences rather than pooling them.

Predictions carry two separate numbers. Probability is the chance the outcome occurs. Confidence is how reliable the prediction itself is.

Highest-weight calls:

• Kalshi asks the appeals court to pause the ban again before August 26, 2026 — 88–94%, High confidence

• Washington receives Kalshi's implementation evidence or a sworn explanation by September 2, rather than silence — 86–94%, Moderate-High

• The next state order containing the information-sharing term names more states than Michigan and Nevada — 85–92%, High

• At least one state not currently suing Kalshi takes formal action by February 28, 2027 — 82–90%, High

• No lawsuit with five or more state plaintiffs emerges before June 30, 2027 — 80–90%, High

• Any new state blocking order names the same vendor or copies the cross-state reporting term — 80–90%, High

• Kalshi obtains no state gambling license anywhere before June 30, 2027 — 74–86%, High

• A state already suing imports two or more Washington-Nevada terms by December 31, 2026 — 72–84%, Moderate-High

• The Commission takes no emergency action directed at Kalshi's compliance with the Washington injunction before October 31, 2026 — 72–82%, Moderate

• The Justice Department files no new preemption suit against Washington before December 31, 2026 — 68–80%, Moderate

Six candidate forecasts remained blocked because the record could not support release. The full analysis publishes those blocked questions alongside the released predictions. The most consequential: nobody has confirmed whether Kalshi met its August 12 compliance deadlines in Nevada and Michigan. No regulator statement, no court filing, no vendor confirmation.

The primary execution found no stable strategic endpoint within the forecast window. Every actor retained at least one profitable move capable of changing the governing regime. The released predictions therefore cover defined events and deadlines rather than a terminal litigation equilibrium.

What Changes for Kalshi

Kalshi's strongest remaining argument is not federal preemption. Preemption has divided courts nationwide and no appellate judge resolves it on an emergency motion.

Kalshi's strongest argument is a drafting gap. The Washington judge listed the categories he found illegal, and technology and science was not among them. The order bans that category anyway. Attacking the gap requires no federal ruling and supports partial rather than total relief.

Expect Kalshi to shift from arguing that location blocking is prohibitively expensive — a position its own Nevada agreement undercuts — toward arguing about scope and legal characterization: prospective access versus exchange regulation, new positions versus cancelled trades, one state's consent versus a transferable admission.

Kalshi also faces a disclosure decision by roughly August 19. The court refused to seal two exhibits supporting Kalshi's opposition, filed in a record that addressed compliance costs and lost profits. Kalshi must file the materials openly or seek passage-specific redactions.

The exhibits may reveal blocking capability, integration requirements, costs, or projected losses. Their actual contents remain unknown until Kalshi files. Separately, the sworn-explanation option in the Washington order would put implementation status on the record, and because that option concerns GeoComply implementation specifically, the cross-state reporting term reaches it.

What Changes for States

States have learned they need not win the national question before acting. Three moves produce working relief.

Read the state's own statutory exemptions narrowly, so a federally licensed exchange does not slip outside the gambling statute. Sue under laws the state already has, including consumer protection statutes that reach false claims about legality. Adopt compliance terms Kalshi has already accepted somewhere else, which costs nothing to draft and is difficult to attack.

Two kinds of state now approach the litigation differently. States banning online sports betting invoke categorical illegality. States licensing sportsbooks invoke unlicensed operation, age restrictions, taxation, integrity controls, and competitive parity. Michigan demonstrates that licensing sports wagering does not eliminate the sports theory — both routes work, and the drafting is simpler for a banning state.

Baltimore's mayor and city council filed suit on August 13 under a municipal consumer protection ordinance, naming Kalshi, Polymarket, and three distribution partners. City governments represent a fourth instrument class, and the filing arrived too late to enter this analysis.

What Changes for Federal Regulators

Michigan is the only state whose order provoked an emergency federal intervention, and the case shows the limits of what the agency can do. Michigan required Kalshi to cancel and refund completed trades. The agency stayed Kalshi's emergency rule and directed normal fulfillment of the open trades, leaving Kalshi caught between a state judge and its regulator.

The agency directs its licensee, stays exchange rules, intervenes, and sues. No federal agency sits above a state court and vacates that court's injunction — a distinction that shapes every forecast about federal response below.

State coercion of a licensed exchange is the foundation of the agency's public case. A settlement the exchange signed voluntarily removes the strongest version of that foundation. Washington's compliance terms reproduce provisions Kalshi accepted in the Nevada joint stipulation, which lowers federal exposure, while the ban itself came from the court and reaches elections and politics — categories the agency defends most aggressively.

The efficient federal strategy is forum selection: pursue rulings that protect market operations or produce binding appellate doctrine, rather than contesting the configuration of every state's blocking technology.

What Remains Unknown

Six facts stay unestablished, and naming them is how the forecast stays honest.

• Whether Kalshi met the August 12 deadlines in Nevada and Michigan

• Which party proposed Washington's compliance terms, which a ten-page joint submission would settle

• The contents of the two exhibits Kalshi unsuccessfully sought to seal

• Michigan's reported $500,000 daily penalty, which appears in press coverage but not in any produced order

• The Sixth Circuit's pending decision in appeals argued July 30

• How Baltimore's municipal theory spreads to other cities

Why Prediction-Market Litigation Is the Hardest Available Test

Most complex litigation resists forecasting for the same reasons it resists management: too many forums, too many decision-makers, and no reliable way to check whether a prediction was right. Prediction-market litigation carries every one of those difficulties and, unusually, makes them public and dated.

Six features make the dispute a working laboratory rather than a news story.

Authority Is Genuinely Contested, Not Merely Disputed. In ordinary litigation, one court decides. Here a federal agency, four federal appeals courts, a dozen state trial courts, state appellate courts, tribal governments, city governments, and state legislatures all hold real instruments. No participant can end the contest alone.

The Rules Can Change During Play. The federal agency is simultaneously a litigant, a friend-of-the-court filer in cases it did not bring, and a rulemaker proposing to define the very term the litigation turns on. A finalized definition would replace the game while the game is being played.

Clocks Run at Different Speeds. State trial courts move in weeks. Federal appeals take months. Rulemaking takes quarters. Legislation takes years. Strategy in a multi-clock system is mostly a question of which deadline arrives first, and that is a forecastable question.

Strategies Are Copyable and the Copying Is Visible. When one state's remedy works, the next state's filing shows whether it borrowed. Diffusion across jurisdictions is directly observable in public dockets rather than inferred from outcomes.

Public and Private Enforcement Run in Parallel. Attorney general actions, gaming-regulator proceedings, municipal ordinances, class actions, and tribal compact claims proceed at once, each with different remedies and different survival conditions.

Outcomes Settle Publicly and on a Date. A docket entry either exists or does not. A deadline is either met or missed. Forecasts about this litigation can be scored against a public record, which is rare in legal analysis and is the precondition for any honest track record.

Most high-stakes commercial disputes share the first five features. Few make the sixth available. Prediction-market litigation therefore functions as a proving ground for forecasting methods intended to travel — to merger review, patent licensing, mass tort, sanctions exposure, and AI governance, where the same multi-forum dynamics operate behind confidentiality.

A Prediction Registry for Prediction-Market Litigation

The subject matter and the method share an epistemology, and MindCast built the analysis to match. Kalshi sells contracts that pay out when a specified real-world event occurs. MindCast publishes forecasts that settle against a specified real-world docket. Both are worthless without a settlement rule stated in advance.

Every released call in the National Prediction Market Litigation Architecture carries four required elements:

• A probability band — how likely the outcome is

• A confidence tier — how reliable the prediction itself is, on a named five-tier scale. A likely outcome resting on thin evidence carries low confidence, and neither number substitutes for the other

• A falsifier — the observation that would prove the call wrong

• A settlement source — the docket, filing, or official release that decides it

Three disciplines keep the registry from becoming advertising.

Blocked Candidates Stay Visible. Six candidate forecasts in the Washington run remained blocked because the record could not support release. Each appears alongside the released predictions with the reason stated, preserving the complete forecasting record.

Corrections Are Recorded as Corrections. The Washington analysis records that a prior MindCast assessment missed one of Washington's three statutory claims entirely, and that every outlet including MindCast carried an August 5 entry date for an order entered August 12. Both entries sit in the validation record next to the calls that landed.

Independent Execution, With Differences Preserved. The Washington forecast ran twice on separate engines. Both agreed on the governing mechanism and differed on route probabilities, and the difference is reported as genuine uncertainty rather than resolved by averaging. Agreement across models strengthens coherence without independently validating a result, since shared assumptions can produce correlated error.

What the discipline produces is not certainty. What it produces is a record an outside party can audit — which forecast was made, when, at what probability, against what falsifier, and whether it came true.

MindCast AI LLC is a predictive behavioral economics and game theory firm. The National Prediction Market Litigation Architecture series tracks federal-state conflict over prediction markets in real time, scores its own prior forecasts against outcomes, and publishes the predictions that fail alongside those that succeed.